Altered Lateral Prefrontal Cortex Functioning during Emotional Interference Resistance is Associated with Affect Lability in Adults with Persisting Symptoms of ADHD from Childhood

Abstract

Background:

Attention-Deficit/Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder characterized by inattention and/or impulsivity/hyperactivity. ADHD, especially when persisting into adulthood, often includes emotional dysregulation, such as affect lability; however, the neural correlates of emotionality in adults with heterogenous ADHD symptom persistence remain unclear.

Methods:

The present study sought to determine shared and distinct functional neuroanatomical profiles of neural circuitry during emotional interference resistance using an Emotional Faces N-Back task while adult participants with persisting (n = 47), desisting (n = 93) or no childhood ADHD symptoms (n = 42) underwent functional magnetic resonance imaging (fMRI) scans.

Results:

Participants without any lifetime ADHD diagnosis performed significant better (faster and more accurately) than did participants with ADHD diagnoses on trials with high cognitive loads (2-back) that included task-irrelevant emotional distractors, tapping into executive functioning and emotion regulatory processes (EF+/ER+). In participants with persisting ADHD symptoms, more severe emotional symptoms were related to worse task performance. Heightened dorsal and ventral lateral prefrontal cortex (dlPFC, vlPFC) activation was associated with more accurate and faster performance on EF+/ER+ trials, respectively. Reduced activation was associated with greater affect lability in adults with persisting ADHD and dlPFC activation mediated the relationship between affect lability and task accuracy.

Conclusions:

These findings suggest that alterations in dlPFC function associated with greater interference in cognitive processes from emotion could represent a marker of risk for problems with emotional dysregulation in individuals with persisting ADHD and thus represent a potential therapeutic target for those with greater emotional symptoms of ADHD.

Introduction

Attention-Deficit/Hyperactivity Disorder (ADHD) is a prevalent, heterogeneous neurodevelopmental disorder characterized by inattention and/or impulsivity-hyperactivity (1). Current diagnostic criteria emphasize cognitive symptoms of the disorder; however, emotional symptoms (e.g., affect lability, anger-irritability) are highly prevalent in both children (25–45%) and adults (30–70%) presenting with ADHD (2). Notably, emotional symptoms in ADHD confer an additional risk for co-morbid psychiatric disorders (e.g., depression), further impair functionality, and are associated with worse clinical outcomes than in cases without such symptoms (3–5). Emotional symptoms can be used to predict patient outcomes over and above “classic” symptoms of inattention and impulsivity/hyperactivity, but the confluence of various cognitive and affective dimensions also contribute to the complex clinical heterogeneity of ADHD (6–8). Despite burgeoning evidence centering emotional symptoms as common features of ADHD etiology (9; 6), our understanding of the neurobiological mechanisms and processes underpinning the affective dimensions of ADHD is sparse, particularly in adult populations. This study aimed to address this gap by elucidating patterns of neural functioning in emotion regulation circuitry across individuals with varying levels of symptom persistence into adulthood to deepen our understanding of the neural pathoetiology of ADHD.

One primary cognitive process of voluntary emotion regulation involves the ability to modulate attention toward or away from emotional information (10,11). ADHD is characterized by difficulty with modulating attention in the context of goal-directed behavior. Evidence suggests that participants with ADHD, relative to those without, perform significantly worse on cognitive (e.g., working memory) tasks when instructed to ignore emotional distractors (i.e., task-irrelevant, affectively charged interference) (12,13). For instance, Marx and colleagues showed that participants with ADHD exhibit emotional interference control deficits even for “low salient” (less charged) emotional distractors, whereas participants without ADHD exhibit performance deficits only for highly charged emotional stimuli, suggesting a lower threshold for affective distractibility in ADHD (12). One such cognitively demanding task with emotional distractors is the Emotional Face N-Back (EFNBACK) task, which has been used to test cognitive-affective symptoms across psychiatric disorders (e.g., depression, bipolar disorder, and pediatric ADHD) while participants undergo fMRI scanning (14–16). Briefly, the task features a low and high cognitive load (either a 0-back or 2-back condition, respectively), and four types of distractors in the form of emotional faces (details in Methods below). The EFNBACK task is thus well-suited to examine the functioning of neural circuitry implicated in voluntary emotion regulation as it relates to emotional symptoms of ADHD.

Brain structures involved in the detection of emotional information—henceforth the “affective salience network”—include the amygdala—a complex of distinct yet inter-connected nuclei involved in salience detection and valence encoding (17,18)—as well as cortical structures, such as the rostral anterior cingulate cortex (rACC) and the ventromedial prefrontal cortex (vmPFC), which facilitate evaluation of emotional stimuli (11). Interactions between these medial frontal structures and the amygdala (sometimes referred to as mPFC-amygdala circuitry) are centrally implicated in emotion regulatory processes in humans and animal models and are disrupted in affective psychopathological states characterized by internalizing symptoms (19–22). Whereas mPFC-amygdala circuitry is involved in aspects of emotion information detection, valence assignment, and rudimentary processing, additional cognitive control regions are thought to downregulate over-excited subcortical regions, dampening impulsivity (via striatal connections) (23,24) and affective reactivity (via amygdalar connections) (25,26). Specifically, lateral PFC regions—like the dorsolateral and ventrolateral prefrontal cortices (dlPFC and vlPFC, respectively)—modulate reactive and/or habitual responses via effortful control by inhibiting prepotent responses and directing and sustaining attention on goal-relevant information (e.g., away from task irrelevant distractors) (26,27).

Neuroimaging studies have found that directing attention away from negatively valenced emotional distractors (e.g., angry or sad faces) compared to neutral distractors (e.g., blank expression faces), for instance, is associated with heightened activation in the affective salience network and reduced activation in lateral PFC regions, and weaker fronto-amygdala functional connectivity (16,29). Researchers have interpreted these findings to indicate that attentional resources dedicated to performing the working memory task and supported by lateral PFC regions are momentarily commandeered by medial frontal and/or temporal structures responding to sensory information that draws attention toward the (often salient and biologically imperative) emotionally distracting stimuli (30). Consistent with this interpretation are findings showing that greater affective salience network activation and reduced lateral PFC (e.g., dlPFC) engagement is associated with worse cognitive performance (i.e., less accurate working memory) (30–35). In contrast, greater vlPFC recruitment has been linked to reduced subjective distractibility and emotionality felt toward negative affective stimuli (33), as well as better working memory performance in the presence of emotional interference 30,31,33,34). Critically, despite evidence that emotional symptoms of ADHD are linked with worse outcomes, few neuroimaging studies have examined the neural substrates of emotion regulation associated with emotional symptoms of ADHD in adults; this is likely due to the historical emphasis of studying ADHD as a developmental disorder.

The present study examined patterns of neural activation in adults with childhood ADHD diagnoses (and either persisting, “ADHD-P” or desisting symptoms “ADHD-D”) and adults without any lifetime ADHD diagnosis (“ADHD-NA”) while undergoing fMRI scanning and performing the EFNBACK task. Our primary hypotheses focused on the 2-back trials with emotional faces as distractors (EF+/ER+) to examine effortful cognitive control during emotional interference, which recruits both executive functioning (EF+) and emotional regulatory (ER+) processes. First, we hypothesized that more severe ADHD symptoms (ADHD-P > ADHD-D/ADHD-NA) would be associated with worse overall performance in terms of task accuracy. Second, we hypothesized that heightened activation in affective salience regions (i.e., amygdala, rACC, vmPFC) and reduced cognitive control regional activation (i.e., dlPFC, vlPFC) would be associated with poor EFNBACK accuracy. Third, we expected to observe significantly greater recruitment of affective salience regions and significantly less engagement of cognitive control regions in the ADHD-P group compared to the ADHD-D and ADHD-NA groups during EF+/ER+ trials. Finally, in the ADHD-P group, who we expect will exhibit higher emotional symptoms, we hypothesized that heightened activation in affective salience regions and/or reduced cognitive control regional activation during 2-back condition with emotionally salient distractors (EF+/ER+) would be associated with more affect lability. To additionally test the extent to which functional activation of the relevant neural substrate explained the association between emotional symptoms and task performance, we conducted an exploratory statistical mediation analysis. That is, although previous work has linked emotional symptoms and cognitive performance in ADHD, we wanted to examine the extent to which activation in implicated neural regions explained the relationship between emotional symptoms and EFNBACK task performance in adults with persisting ADHD (ADHD-P).

Methods

2.1. Participants

The starting sample in the present study consisted of 256 participants who were recruited into the study from the Pittsburgh ADHD Longitudinal Study (PALS), a sample diagnosed with ADHD (per DSM-III-R or DSM-IV criteria) as children between 1987 and 1996 using comprehensive, standardized, multi-informant diagnostic methods including clinician consensus (36,37). Of those, 182 provided usable fMRI data (criteria described below and in the Supplement), which included participants with persisting ADHD symptoms (ADHD-P, n = 47; mean age = 34.74 ± 3.87; 85.1% male), those with desisting ADHD symptoms (ADHD-D, n = 93; mean age = 34.81 ± 3.26; 93.5% male), and participants without a childhood ADHD diagnosis (ADHD-NA, n = 42; mean age = 35.57 ± 3.75; 85.7% male). See Table 1 for sample characteristics.

Table 1.

Demographic Characteristics of Participants

Clinical Measures		n	%
Sex, Male		163	89.6%
Race
	American Indian or Alaskan Native	2	1.1%
	Asian	1	0.6%
	Native Hawaiian or other Pacific Islander	0	0%
	Black or African American	24	13.2%
	White	141	77.5%
	More than one race	11	6.0%
	Unknown or not reported	3	1.7%
Ethnicity
	Hispanic or Latino	2	1.1%
	Not Hispanic or Latino	166	91.2%
	Unknown or not reported	14	7.7%
Education
	Less than high school	1	0.6%
	Attended high school but did not graduate	3	1.7%
	High school graduate or GED equivalent	29	15.9%
	Completed technical/secretarial or other specialized training	13	7.1%
	Partial college (at least one year)	29	16.0%
	Associate’s or 2 year degree	31	17.0%
	College or university graduate	49	26.9%
	Graduate or professional training (graduate degree)	27	14.8%
Monthly Income
	$0	9	4.9%
	Less than $200	4	2.2%
	$200-$499	5	2.7%
	$500-$999	20	11.0%
	$1,000-$1,999	30	16.5%
	$2,000-$2,999	37	20.3%
	$3,000-$4,999	34	18.7%
	$5,000-$6,999	22	12.1%
	$7,000 or more	21	11.5%
ADHD Symptom Persistence
	Persisting	47	25.8%
	Desisting	93	51.1%
	No ADHD History	42	23.1%

Exclusion criteria for the present study included MRI scanning contraindications (e.g., non-removable metal, claustrophobia), diagnoses of neurological disorders (e.g., seizures, meningitis, or encephalitis) or conditions (e.g., concussion with loss of consciousness > 5 minutes), certain psychiatric disorders (e.g., schizophrenia, psychosis, severe substance use disorder—per DSM-5—excluding tobacco), certain medications (e.g., blood pressure medications), and weight > 300 lbs. Participants prescribed stimulant medications (n = 11) were required to refrain for 24 hours prior to the MRI scan session and participants who smoke were asked to abstain for 2 hours prior to the scan (see Supplement for tobacco use breakdown). Additional details on the PALS neuroimaging sample can be found in previous publications (38,39).

2.2. Procedure

After obtaining informed consent and screening for drugs/alcohol using saliva-based testing, participants practiced scanner tasks in an MRI simulator to familiarize themselves with the neuroimaging protocol (see Supplement for substance use breakdown). Following an out-of-scanner practice session, participants completed the neuroimaging protocol, cognitive tasks, and questionnaires (details below). This research was approved by the University of Pittsburgh Human Research Protections Office.

2.3. Neuroimaging Data

Acquisition

Neuroimaging data were acquired on a 3 Tesla MRI scanner that underwent an upgrade from a Siemens Trio to a Siemens Magnetom Prisma (Siemens 3T Trio, n = 116; Siemens 3T Prisma, n = 66). We harmonized the datasets using NeuroComBat to account for collecting neuroimaging data on an MRI scanner before and after the upgrade and included ADHD group membership (i.e., ADHD-P, ADHD-D, ADHD-NA) as a covariate to pool data while preserving relevant biological variability of interest (40). See Supplement for MRI scan acquisition details and preprocessing procedures.

ROI definitions

Regions-of-interest (ROIs) in cortico-amygdala regions were chosen to represent two aspects of emotional regulatory processes involved in resisting valenced distractors; these included top-down control regions in the lateral PFC (specifically, dlPFC and vlPFC) as well as structures constituting an affective salience network, which included the vmPFC, rACC, and the centromedial and basolateral amygdala. All six ROIs were defined using the Brainnetome atlas and included both left and right hemispheres (41). See Supplement for specific ROI definitions and Figure 1 for an illustration of the ROIs on a standard brain template.

Figure 1.

Regions-of-Interest (ROIs)

Regions-of-interest (ROIs) defined using the Brainnetome atlas. Abbreviations: vmPFC, ventromedial prefrontal cortex; rACC, rostral anterior cingulate cortex; BLA, basolateral amygdala; CMA, centromedial amygdala; dlPFC, dorsolateral prefrontal cortex; vlPFC, ventrolateral prefrontal cortex.

2.4. Emotional Face N-Back (EFNBACK) task

The EFNBACK task is a modified visual sequential letter working-memory N-back task with emotional faces presented as distractors (42). As a neuroimaging task, the EFNBACK has elicited neural activation in regions associated with emotion regulation in several neuropsychiatric disorders (14,42, 43), underscoring its ability to discriminate functional differences across psychopathologies. See Supplement for task details. Given our interest in neural systems of regulation in the context of emotional interference resistance in adult ADHD probands, we focused our analyses on the conditions that included (neutral, negative, or positive) faces. We refer to this combination of trials, which include high cognitive and affective loads, as ‘EF+/ER+’ to identify these trials as aimed at engaging processes related to executive functioning and emotion regulation, respectively. Our two EFNBACK behavioral measures of interest were accuracy—defined as the percent of total correct trials—and response times—defined as the length of time (in milliseconds) participants took to respond on correct trials. See Figures 2A and B for distributions of EFNBACK accuracy and response times, respectively.

Figure 2.

Behavioral Task Performance and Affect Lability Scores

Distribution of EFNBACK task trial mean accuracy (A) and response times on correct trials (B), along with distribution of ALS score (C) for each group. Abbreviations: EFNBACK, Emotional Face N-Back; EF-, 0-back no emotional faces trials; EF+/ER+, 2-back emotional faces trials; ms, milliseconds; ADHD-P, adults with persisting ADHD symptoms; ADHD-D, adults with desisting ADHD symptoms; ADHD-NA, adults with no history of ADHD symptoms; ALS, Affect Lability Scale.

2.5. Cognitive and Emotional ADHD Symptomatology

ADHD (cognitive) symptoms during adulthood were assessed using mean scores from the Barkley Adult ADHD Rating Scale IV (BAARS-IV) (44). As described in previous publications using this sample (38,39), the BAARS-IV is an 18-item questionnaire with responses ranging from 0–3. ADHD symptoms were persistent if five or more symptoms of inattention or impulsivity-hyperactivity were present, consistent with DSM-5 criteria. Symptoms were considered “present” by taking the higher response from either self- or collateral informant-report to address potential underreporting. BAARS-IV scores had excellent internal reliability (Cronbach’s α = .95). We used mean scores from the Affective Lability Scale 18-item form (ALS-18) (45) to assess shifts in mood on a 4-point Likert scale, ranging from “Very uncharacteristic of me” to “Very characteristic of me.” ALS scores had excellent internal reliability across the entire sample (Cronbach’s α = .93). See Figure 2C for distributions of ALS scores by participant group.

Results

3.1. Task effects

3.1.1. Behavioral

There was no significant condition-by-group interaction effect on mean accuracy (F = 2.226, p = .109, η² = .007) but there were significant main effects of both group (F = 8.469, p < .001, η² = 0.026) and condition (F = 30.465, p < .001, η² = 0.046) on accuracy. Post-hoc pairwise t-tests revealed that participants in the ADHD groups were significantly less accurate than those in the ADHD-NA group (ADHD-P: Bonferroni p < .001; ADHD-D: Bonferroni p = .017) across conditions. There was no significant difference in trial accuracy across conditions between the ADHD-P and ADHD-D groups (Bonferroni p = 1.0). Post-hoc tests also revealed that across groups, participants were significantly more accurate on EF- than on EF+/ER+ trials (p < .001).

There was no significant condition-by-group interaction effect on response times for correct trials (F = .697, p = .499, η² = 0.002). There was also no significant main effect of group on correct-trial response times (F = 2.976, p = .052, η² = 0.009) but there was a significant main effect of condition (F = 90.193, p < .001, η² = 0.126). Post-hoc tests revealed that participants across groups were significantly slower on correct EF+/ER+ trials than on EF- trials (p < .001).

3.1.2. fMRI

There was no significant group-by-condition interaction effect on activation in any of the ROIs examined (ps ≥ .395). There was a significant main effect of group on activation in the basolateral amygdala ROI (F = 4.144, p = .016, η² = 0.006) but no other region (ps ≥ .117); however, a post-hoc pairwise t-test revealed no significant between group differences in basolateral amygdala activation (ps > .074). There were also significant main effects of condition on activation in the dlPFC (F = 16.675, p < .001), vlPFC (F = 32.458, p < .001, η² = 0.022), basolateral amygdala (F = 4.042, p = .045, η² = 0.003), rACC (F = 11.519, p < .001, η² = 0.008), vmPFC (F = 66.941, p < .001, η² = 0.045) but not in the centromedial amygdala (p = .077). Post-hoc tests revealed significantly elevated activation across participants on EF+/ER+ trials than on EF- trials in both top-down control network ROIs: dlPFC (p = .026) and vlPFC (p = .0003). In contrast, post-hoc tests revealed significantly reduced activation on EF+/EF+ trials than on EF-trials in several affective salience network regions: basolateral amygdala (p = .012), rACC (p = .012), and vmPFC (p < .001).

See Figure 3 for distributions of fMRI BOLD task activation betas by ROI.

Figure 3.

fMRI Task BOLD Activation

Distributions of EFNBACK fMRI task BOLD activations (beta) on (a) EF+/ER+ and (b) EF-trials across cortico-amygdala ROIs separated by group. Abbreviations: BOLD, blood oxygen level dependent; EFNBACK, Emotion Face N-Back; EF+/ER+, 2-back emotional faces trials; dlPFC, dorsolateral prefrontal cortex; BLA, basolateral amygdala; CMA, centromedial amygdala; rACC, rostral anterior cingulate cortex; vlPFC, ventrolateral prefrontal cortex; vmPFC, ventromedial prefrontal cortex; ADHD-P, adults with persisting ADHD symptoms; ADHD-D, adults with desisting ADHD symptoms; ADHD-NA, adults with no history of ADHD symptoms.

3.2. Associations between Neural Activation and EFNBACK Performance

Given our interest in understanding resistance to emotional interference, we focused our primary analyses examining the relationship between neural activation and EFNBACK performance on EF+/ER+ trials.

3.2.1. Accuracy

There were significant group-by-condition interactions on EFNBACK EF+/ER+ trial accuracy in the dlPFC (F = 10.235, p < .001, η² = 0.019), vlPFC (F = 7.810, p < .001, η² = 0.014), and centromedial amygdala (F = 3.909, p = .020, η² = 0.007) but not in any other ROI (ps ≥ .227). Post-hoc tests revealed that heightened activation in all three ROIs was associated with higher accuracy on EF+/ER+ trials in the ADHD-P group: dlPFC (F = 21.907, p < .001, η² = 0.073), vlPFC (F = 11.805, p < .001, η² = 0.041), and centromedial amygdala (F = 6.336, p = .012, η² = 0.022). In contrast, heightened activation in all three ROIs was associated with lower accuracy on EF+/ER+ trials in the ADHD-NA group: dlPFC (F = 4.656, p = .032, η² = 0.018), vlPFC (F = 13.538, p < .001, η² = 0.052), and centromedial amygdala (F = 4.459, p = .036, η² = 0.018). Finally, in contrast to both the ADHD-P and ADHD-NA groups, neural activation in these regions was not associated with EF+/ER+ trial accuracy in the ADHD-D group: dlPFC (F = .009, p = .926), vlPFC (F = 2.394, p = .122), and centromedial amygdala (F = .007, p = .934).

See Figure 4A for associations between ROI activation and EFNBACK EF+/ER+ accuracy by group.

Figure 4.

Lateral PFC Activation on EF+/ER+ Trials and Task Performance

Associations between lateral PFC BOLD activation during EF+/ER+ trials and EFNBACK mean accuracy (A) and response times on correct trials (B) for each group. Abbreviations: BOLD, blood oxygen level dependent; PFC, prefrontal cortex; BOLD, blood oxygen level dependent; EF+/ER+, 2-back emotional faces trials; dlPFC, dorsolateral prefrontal cortex; vlPFC, ventrolateral prefrontal cortex; ADHD-P, adults with persisting ADHD symptoms; ADHD-D, adults with desisting ADHD symptoms; ADHD-NA, adults with no history of ADHD symptoms.

3.2.2. Response Times

There were significant group-by-condition interactions on EFNBACK EF+/ER+ correct-trial response times in the dlPFC (F = 8.117, p = .003, η² = 0.015), vlPFC (F = 13.364, p < .001, η² = 0.024), rACC (F = 11.831, p < .001, η² = 0.022), and basolateral amygdala (F = 3.826, p = .022, η² = 0.007) but not in the vmPFC or centromedial amygdala (ps ≥ .383). Post-hoc tests revealed that activation in neither top-down control ROIs was associated with response times in the ADHD-P group (both ps ≥ .202); however, heightened activation in both regions was associated with slower response times in the ADHD-D group (dlPFC: F = 6.154, p = .013, η² = 0.011; vlPFC: F = 16.125, p < .001, η² = 0.029) and ADHD-NA group (dlPFC: F = 27.628, p < .001, η² = 0.10; vlPFC: F = 37.788, p < .001, η² = 0.132). In the ADHD-P group, heightened rACC activation was associated with faster response times on correct ER+/EF+ trials (F = 4.99, p = .026, η² = 0.018), whereas heightened rACC activation was associated with slower response times in both the ADHD-D (F = 10.374, p = .001, η² = 0.019) and ADHD-NA (F: 18.576, p < .001, η² = 0.07) groups on correct EF+/ER+ trials. Basolateral amygdala activation was not associated with response times in either ADHD group (both ps ≥ .122) but heightened activation in this region was associated with shorter response times in the ADHD-NA group (F = 5.321, p = .020, η² = 0.021).

See Figure 4B for associations between ROI activation and EFNBACK EF+/ER+ response times on correct trials by group.

3.3. Associations with Affect Lability

An ANOVA test revealed significant group differences in emotional lability F(2, 253) = 15.13, p < .001. Post-hoc pair-wise t-tests revealed significantly greater emotional lability in the ADHD-P group compared to the ADHD-D group (p < .001) and ADHD-NA group (p < .001). However, following multiple comparisons correction, the ADHD-D group did not significantly differ in emotional lability from the ADHD-NA group (p = .062).

Given our interest in understanding emotional symptoms in adults with persisting ADHD symptoms, we performed analyses to test the relationship between EFNBACK performance and neural activation with affect lability in the ADHD-P group on EF+/ER+ trials.

3.3.1. Behavioral

Greater affect lability was associated with lower accuracy on EF+/ER+ trials in ADHD-P participants (F = 14.558, p < .001, η² = 0.095). Affect lability was not associated with response times on correct trials in the ADHD-P group (p = .947).

3.3.2. fMRI

Greater affect lability was associated with reduced activation in top-down control regions in the ADHD-P group (dlPFC: F = 18.468, p < .001, η² = 0.063; vlPFC: F = 23.356, p < .001, η² = 0.078); however, affect lability was not associated with activation in affective salience network ROIs (ps ≥ .442).

3.4. Exploratory Mediation Analysis

Exploratory mediation analyses tested the extent to which activation in top-down control ROIs (i.e., dlPFC, vlPFC) during EF+/ER+ trials mediated the association between affect lability and EFNBACK accuracy. Our analysis controlled for mean scores of overall ADHD symptom severity to ensure that any detected effects were specific to emotional symptoms and not accounted for by cognitive aspects of ADHD (e.g., inattention and/or impulsivity-hyperactivity).

We found that dlPFC activation during EF+/ER+ partially mediated the association between affect lability and EFNBACK accuracy. The results of the mediation analysis after bootstrapping revealed a significant indirect (mediating) effect of dlPFC activation during EF+/ER+ trials in the relationship between affect lability and EFNBACK accuracy (ab; β = −.01, p = .004), with a significant proportion of the relationship mediated by the dlPFC (β = .16, p = .004). There was also a significant direct effect (c’; β = −.04, p < .001), as well as a significant total effect (c; β = −.05, p < .001). In sum, results indicated that affect lability had both direct and indirect effects on (associations with) EFNBACK accuracy through dlPFC activation (see Figure 5 for mediation model), above and beyond ADHD symptom severity. Critically, this test represents a statistical mediation, as causal mediation cannot be established with the present cross-sectional study design.

Figure 5.

Mediation Model

Exploratory mediation model showing dlPFC activation during EF+/ER+ trials mediating the association between affect lability and task accuracy, after controlling for ADHD symptoms in adults with persisting ADHD (ADHD-P; n=47). Abbreviations: dlPFC, dorsolateral prefrontal cortex; EF+/ER+, 2-back emotional faces trials; EFNBACK, Emotional Faces N-Back; ADHD, Attention-Deficit Hyperactivity Disorder. *p < .05, ** p < .01, *** p < .001. The a path is the association between affect lability and the mediator variable (dlPFC activation). The b path is the association between the mediator variable and EFNBACK EF+/ER+ trial accuracy. The c’ path is the direct effect of affect lability on EFNBACK EF+/ER+ trial accuracy. The c path is the total effect of affect lability on EFNBACK EF+/ER+ trial accuracy.

Discussion

Emotional symptoms are common in adults with persisting ADHD symptoms, yet their functional neuroanatomy remains poorly understood. In the present study, we used an emotional face working memory paradigm (EFNBACK) while adult participants with persisting, desisting, or no childhood or adulthood ADHD symptoms underwent fMRI scanning. Findings show that reduced activations in the dlPFC and vlPFC during high cognitive load (2-back) and emotional distracters (trials with faces) in adults with persisting ADHD symptoms were associated with more severe affect lability. This association remained significant even after accounting for variability in inattention and impulsivity/hyperactivity symptoms within the ADHD symptom persistent group, suggesting that emotional symptoms may correlate with additional behavioral and neural variability. Mediation analyses further showed that dlPFC activation during EF+/ER+ trials in the ADHD-P group statistically mediated the association between affect lability and performance (task accuracy), above and beyond ADHD symptoms. Taken together, our results implicate the dlPFC—a cortical structure involved in attention selection and cognitive control—in emotional dysregulation in adults with persisting ADHD, such as in affect lability.

Our findings largely support our main hypothesis: as expected, less activation in lateral PFC regions was associated with more emotionality (e.g., affect lability); however, we did not observe a positive relationship between activation in affective salience structures (e.g., the amygdala or vmPFC) and affect lability as we had expected. In addition, although greater activation in the dlPFC and vlPFC was associated with better performance on EF+/ER+ EFNBACK trials, this association was only present in the ADHD-P group. Surprisingly, stronger vlPFC activation was associated with worse performance in the ADHD-NA group, suggesting that the structure may differentially relate to behavioral performance on a cognitive task in adults with versus without ADHD.

Previous research using the EFNBACK task has documented similar working memory deficits in adults with, compared to adults without ADHD, along with more pronounced difficulties with emotional interference control in the former group relative to the latter 12). Earlier research has implicated the ACC in adults with ADHD when viewing negative versus neutral images (46); however, this work contrasted negative and neutral pictures to test emotional processing and regulation, to which our results cannot directly speak. A large body of evidence also implicates the amygdala and vlPFC on working memory performance when facing emotional interference (33,29), which is not necessarily at odds with our findings: although our fMRI task included emotional faces, our study design did not include subjective measures of distractibility or eye movement indices of distraction thereby precluding any conclusion that participants experienced the distracting stimuli as salient or emotionally evocative.

Our results differ from previous work that has used the EFNBACK fMRI task in adults with and without ADHD. Albeit with fewer participants, the study investigators failed to observe a significant difference in dlPFC activation between adults with and without ADHD (47). The authors of that study attributed the lack of group difference in dlPFC (and amygdala) activation to the fact that they recruited adult participants, suggesting that symptoms conferring functional impairment present during childhood and into adolescence may no longer be present in older participants. Our study extends this work by recruiting and differentiating adults with persisting from desisting ADHD symptoms at the time of assessment, which may explain some of the discrepancies in findings. Indeed, our results are consistent with previous work performed in younger participants, which has shown dlPFC differences in structure (48) and function (49) in emotional symptoms in participants with ADHD relative to typically developing youth.

Interestingly, heightened activation in the dlPFC and vlPFC during EF+/ER+ EFNBACK trials was associated with more accurate performance on these trials but only in the ADHD-P group. We interpret these findings to suggest that successful lateral PFC recruitment and engagement in adults with persisting ADHD symptoms may be associated with greater emotional interference resistance. This is consistent with our result showing that greater engagement of the dlPFC/vlPFC in adults with persisting ADHD symptoms was associated with less affect lability.

Recent evidence has suggested a complex dynamic between cognitive and affective processes in ADHD, with some suggesting that deficits in working memory and attentional control may contribute directly and indirectly to symptoms characterized by emotion dysregulation (50,51). Researchers investigating emotion regulatory strategies reported that individuals with ADHD may be employing more emotional suppression as a compensatory approach to manage emotionality rather than cognitive reappraisal (52), with the latter being more effective in reducing negative affect relative to the former linked with heightened physiological response 53. Given that current conceptualizations of emotional interference resistance underscore lateral PFC structures (31,32), it is plausible that functional impairment of these neural structures contributes to heightened emotionality (e.g., lability) in ADHD, possibly due to less successful cognitive reappraisal. However, because the present study did not explicitly elicit emotional states or test regulatory strategies, this too would require explicit testing in adults with varying levels of symptom persistence as well as with heterogeneous affective and cognitive control in varying contexts. See Supplement for an extended Discussion and strengths/limitations.

Conclusion

In sum, findings from this study add to the sparse literature on the neural underpinnings of voluntary emotion regulation linked to emotional symptoms in adults with childhood-onset ADHD. Because the adults in our samples were diagnosed in childhood and followed through adulthood, it was possible to examine whether ADHD symptom persistence is associated with alterations in the functioning of emotion regulation circuitry and emotional symptoms of ADHD. Our findings identify reduced functional activation in lateral cortical structures (i.e., dlPFC, vlPFC) as likely contributors to the pathoetiology of ADHD symptoms related to affective control. Given emerging evidence indicating promising results for cognitive symptoms of ADHD when targeting the lateral PFC (e.g., using pharmacological or neuromodulatory approaches), we encourage future investigators to consider measuring changes in the severity of emotional symptoms in adults with ADHD when carrying out such experiments.

Table 2.

Correlation matrices (Pearson’s r values) of variables of interest for EF+/ER+ (Emotional Face N-back) EFNBACK trials in ADHD probands (ADHD-P, ADHD-D).

	2	3	4	5
1. EFNBACK accuracy	−.46 ***	−.002	.01	−.05
2. EFNBACK response time		.07	.06	.01
3. Inattention			.76 ***	.38 ***
4. Impulsivity/hyperactivity				.41 ***
5. Affect lability

‘Accuracy’ refers to the number correct trials over total trials. ‘Response time’ refers to time taken in milliseconds of correct trial responses. ADHD symptoms of inattention and impulsivity/hyperactivity were assessed using the Barkley ADHD scale and affect lability was evaluated using the Affect Lability Scale.

^*p < .05

^**p < .01

^***p < .001